Market Thickness and The Early Labour Market Career of University Graduates

Spatial Economic Analysis, 2014
Vol. 9, No. 4, 396419, http://dx.doi.org/10.1080/17421772.2014.961534
Market Thickness and the Early Labour Market Career

of University Graduates: An Urban Advantage?
LINA AHLIN, MARTIN ANDERSSON & PER THULIN

(Received February 2013; accepted March 2014)
ABSTRACT We analyse the influence of market thickness for skills on initial wages and the early job
market career of university graduates. Using Swedish micro-level panel data on a cohort of graduates,
we show that two out of three graduates move to large cities upon graduation. Large cities increase
employment probabilities and yield higher rewards to human capital, even after controlling for
employment selection. The premium on initial wages for graduates in urban regions is in the interval of
56%, and we estimate a wage-growth premium of about 24%. Thicker markets for skills appear as
a key reason for the concentration of graduates to larger cities.
paisseur du march initial de lemploi pour les diplms universitaires un

avantage en ville?
RSUM nous analysons linfluence de lpaisseur du march en comptences sur les salaires de dpart,
et le march initial de lemploi pour les diplms universitaires. En utilisant des donnes de panel
micro-niveau, en Sude, appliqus une cohorte de diplms, nous montrons que deux diplms sur
trois se rendent dans une grande ville ds quils se sont diplms. Les grandes villes augmentent les
possibilits de trouver un emploi, et apportent une rentabilit suprieure pour le capital humain, mme
en tenant compte du processus de slection. La rmunration de dpart des diplms en zone urbaine
est suprieure de lordre de 5 6 pour cent, et nous estimons que laugmentation des salaires est
suprieure, elle aussi, de 2 4 pour cent. Des marchs plus pais pour les comptences semble tre la
principale raison de la concentration de diplms dans les grandes villes.
Lina Ahlin, Department of Economics, Centre for Innovation, Research and Competence in the Learning
Economy (CIRCLE), Lund University, PO Box 117, S-22100 Lund, Sweden. Email: lina.ahlin@circle.lu.se.
Martin Andersson (to whom correspondence should be sent), Centre for Innovation, Research and Competence in
the Learning Economy (CIRCLE), Lund University, Lund and the Department of Industrial Economics,
Blekinge Institute of Technology (BTH), Karlskrona, Sweden. Email: martin.andersson@circle.lu.se. Per Thulin,
Department of Industrial Economics and Management, Royal Institute of Technology, Stockholm and the
Swedish Entrepreneurship Forum, Lindstedtsvgen 30, S-10044 Stockholm, Sweden. Email: per.thulin@indek.
kth.se. We are grateful for constructive comments from two anonymous reviewers and seminar participants at
CIRCLE, Lund University and the Jnkping International Business School, as well as from participants at the
special session on Graduates, Entrepreneurship and Regional Labor Markets at the ERSA congress in Bratislava,
Slovakia in August 2012, and the WRSA annual meeting 2013 in Santa Barbara, USA. This work has been
supported by financial grants from the Swedish research council FORMAS (Dnr 201180), as well as the Swedish
Research Council [Linnaeus Grant No. 349200680] and the Swedish Governmental Agency for Innovation
Systems [Grant agreement 201007370].
2014 Regional Studies Association
397
La densidad del mercado y el inicio de carrera en el mercado laboral de los

egresados universitarios una ventaja urbana?
RESUMEN Analizamos la influencia de la densidad del mercado de habilidades sobre los salarios
iniciales y el inicio de carrera en el mercado laboral de los egresados universitarios. Utilizando los datos
de un panel sueco de micronivel de un grupo de licenciados, demostramos que dos de cada tres
licenciados se mudan a las grandes ciudades despus de que obtienen su ttulo. Las grandes ciudades
incrementan las probabilidades del capital humano en relacin con el empleo y remuneraciones ms
altas, incluso despus del control para la seleccin del empleo. El mayor pago en relacin con la
remuneracin inicial de los licenciados en las regiones urbanas est en el margen de entre el 5% y 6% y
estimamos un mayor valor del incremento salarial de entre el 2% y 4% aproximadamente. En lo que
respecta a las habilidades, los mercados ms densos parecen ser la razn primordial para la concentracin
de licenciados en las ciudades de mayor tamao.
?
: ,
, ,
,
,
5% 6%, 2%
4% ,

KEYWORDS: Human capital; university graduates; urban wage premium; market thickness; matching;
agglomeration economies; mobility; job switching

JEL CLASSIFICATION: R12; R10; J61; J31
1. Introduction
The spatial distribution of human capital plays a key role in shaping the long-term
geography of jobs, incomes and well-being. Analyses of the US and European
regions show that highly educated workers are important drivers of regional
development in terms of jobs as well as incomes per capita (Glaeser et al., 1995;
Simon, 1998; Cheshire & Magrini, 2000; Badinger & Tondl, 2003). Shapiro (2005)
reports that a 10% increase in a metropolitan regions concentration of workers
with a university education leads to a 0.8% increase in subsequent employment
growth in the USA. Moretti & Thulin (2013) use data for Sweden and find a local
multiplier in the order of 3 for manufacturing jobs associated with a long university
education, i.e. about 6 times higher than for manufacturing jobs in general. Human
capital also consistently comes out as a significant determinant of innovation and
technology absorption in local industries and firms (Faggian & McCann, 2009a;
Andersson & Lf, 2012).
398
L. Ahlin et al.
The general trend is indeed that regions that manage to attract and retain
human capital prosper, whereas others fall behind. Moretti (2012) considers the
Great Divergence taking place between regions with high and low concentrations of highly educated individuals to be one of the most important developments
in the US economy during the last decades. Understanding the location of human
capital is thus a crucial issue.
We focus in this paper on the location of university graduates, i.e. highly
educated new labour market entrants. University graduates are important as they
embody newly acquired knowledge and skills, and their location choice has a
direct influence on the spatial distribution and regional growth of stocks of human
capital (Bradley & Taylor, 1996). They are also mobile in the sense that they need
to make a choice about where to enter the labour market after graduation, and
their decisions are not distorted by any cumulated labour market experiences and
historical career paths.
We set out to understand the preference of university graduates towards
starting to work in large urban regions after graduation. Using data for Sweden, we
document a significant concentration of graduates to the countrys three major
urban regions. Almost two out of three graduates enter employment in large urban
regions after graduation. This number is significantly higher than what is motivated
by the fraction of graduates that graduate from a university in these regions (which
is about 45%), and means that the large urban regions are net receivers of graduates
educated elsewhere. Their attraction is also higher than what is motivated by the
general concentration of workers to these regions (which is about 40%). These
numbers suggest that understanding the location of university graduates in Sweden
is much about understanding their preference for entering the labour market in the
larger urban regions upon graduation.1
There are two main explanations for the concentration of highly educated
workers in larger regions. Authors such as Glaeser & Gottlieb (2006) emphasize the
variety of consumption amenities (such as pubs, restaurants, theatres, specialty
stores, opera houses) offered by larger city regions as a key factor behind these
regions attractiveness for highly skilled workers. In contrast, Scott (2010) as well as
Storper & Scott (2009) claim that amenity-based arguments are weak and argue in
favour of job availability. In an analysis of Dutch graduates, Venhorst et al. (2011)
find support for the labour market argument in that a large labour market is the
most important characteristic of those regions that manage to retain graduates from
their local universities.
Our analysis takes a labour market approach and emphasizes the advantages of
urban regions from the perspective of market thickness for skills. University
graduates are a select group that enter the job market for the first time upon
graduation, and are likely to place a large value on job availability and early job
market career prospects. Large urban cities offer a greater number of jobs associated
with university education and also host greater concentrations of skilled workers.2
Models of the efficiency gains of market thickness suggest that these attributes of
local labour markets increase matching probabilities (Gan & Li, 2004) as well as the
expected quality of each match (Hesley & Strange, 1990; Kim, 1990). University
graduates entering the labour market in larger urban cities are thus more likely to
find a better job that matches their skills set.
But market thickness for skills does not only imply advantages that accrue to
the probability of getting a first quality job but also influences the long-term
reward to human capital and early labour market career development. University
399
graduates with no prior experience of the labour market are likely to have little
knowledge of what kind of work and characteristics of employers they have a
preference for, or simply what they are good at in working life. For new entrants
to the labour market, Topel & Ward (1992) show that the typical path to a stable
employeremployee relationship is characterized by frequent job switching, and
this can be understood as a consequence of a search process for a match of high
quality, where match quality is an experience good (Farber, 1994).3 Through
thicker markets for skills, large urban regions offer better opportunities and lower
transactions costs associated with job switching, and such mobility may be necessary
to find a good job and advance early in the job market career. For these reasons,
university graduates may thus prefer a large region with a variety of employers and
jobs to a small region with a narrower job market even in circumstances when they
have job offers in both regions. Not only are the prospects of a first quality job
better, but more importantly, the prospects of a good job market career over time
rewarding their human capital investments.
We argue that these job market advantages of large cities are an important
reason for their attraction of university graduates. In particular, the described
benefits of thicker markets for skills should be reflected in higher initial wage
earnings as well as higher wage growth during the early job market career for
university graduates entering the labour market in large cities. To test these arguments, we study the link between the location choice of university graduates in
Sweden and their subsequent wage growth and labour market activity in their early
career. Every graduate in 2000 is followed during their first 9 years on the labour
market (20012009). We assess the location choice of university graduates and how
initial wages, wage growth as well as rates of job switching relate to their location
choice while accounting for observed and unobserved individual characteristics.
We show that there is a strong preference of moving to one of the countrys
metropolitan regions after graduation. University graduates that enter the job
market in large cities after graduation are a select group, and appear to be more
motivated as judged by their high-school grades as well as their socio-economic
background. Graduates in urban regions are more likely to find a job and also
experience a stronger early job market career. Controlling for employment
selection, we show that they have higher initial wages and experience higher
wage growth. Our results lend strong support for that thicker markets for skills,
which improves matching and early job market career prospects, are a key reason
for the concentration of university graduates to larger cities.
The rest of the paper is organized as follows: Section 2 discusses previous
literature and theoretical perspectives on the location of university graduates.
Section 3 presents the data, defines variables and provides descriptive statistics. The
general patterns as regards location, wages and job switching among our sample of
university graduates are presented in Section 4. Section 5 presents the empirical
strategy and our results. Section 6 presents the conclusions.
2. Thick Markets for Skills, City Size and Career Opportunities of

University Graduates
Why do university graduates show a strong preference for entering the job market
in large cities? There are several explanations put forth in the literature. The two
major strands of literature either emphasize consumption amenities (Glaeser &
400
L. Ahlin et al.
Gottlieb, 2006) or job availability (Storper & Scott, 2009). The first argument
concerns that skilled individuals choose to locate in regions where there are
amenities such as an abundance of pubs, restaurants and access to fine arts. The
counter argument by authors such as Storper & Scott (2009) is that these amenities
only exist because individuals have already chosen to locate there, thus creating a
market for these amenities, and that there consequently must be some other factors
behind the concentration of workers to big cities. In lieu, economies of scale in
production and advantages of thick markets are considered to drive the
urbanization process in providing a diverse labour market for the highly skilled,
where new sectors are able to provide growth potential (and jobs) if the
dominating sectors decline. In short, individuals locate where the jobs are.
Some evidence in favour of this labour market argument is provided by
Venhorst et al. (2011) for the Netherlands who find support for that a large labour
market is the most important characteristic of those regions that manage to retain
graduates from their local universities. In addition, Faggian & McCann (2006,
2009a) find that only some types of higher education institutions are efficient in
retaining their graduates in the region and that this is largely dependent on the
overall labour market characteristics in the surrounding region. Thus, recent
graduates tend to enter into employment where there is an abundance of jobs
rather than an abundance of amenities, although density of amenities typically goes
hand in hand with density of jobs.
In the Swedish context, there is also recent evidence supporting job-related
arguments. Using survey and register data, Niedomysl & Hansen (2010) find that
jobs rather than amenities are more important for recent university graduates
mobility decision. Furthermore, in a survey of 2,000 highly educated people
belonging to Swedish career network 4Potentials, it is shown that job opportunities, availability of more qualified positions within firms and the possibility of
working internationally are primary reasons for location decisions. Amenities were
only considered secondary (4Potentials, 2012).
Another job-related reason for the increased urbanization of the highly
educated is associated with the increase in highly educated female professionals
over the recent decades. Costa & Kahn (2000) argue that the major driver for
highly skilled couples to locate in metropolitan areas is the prevalence of a diverse
and large labour market where both partners in a dual-career household (deemed
power couples when married to one another) can find suitable jobs. This
colocation problem is something that a large part of the literature on the urban
wage premium neglects since their data are often limited to (white) males with
diverse labour market experiences (cf. Topel & Ward, 1992; Glaeser & Mar,
2001; Wheeler, 2006; Yankow, 2006).4 The analyses that follow include all
university graduates, males as well as females, and we have partnership information
including the education level of the partner.
A common observation is that areas with a high concentration of human capital
provide higher wages, i.e. a so-called urban wage (or productivity) premium which
may influence the localization decision of recent graduates if they locate where
their education gets rewarded. Studies on the urban wage premium have generally
found a nominal wage premium of 26% for workers locating in dense urban areas
(Glaeser & Mar, 2001; Yankow, 2006; Wheeler, 2006 for the USA; Lehmer &
Mller, 2010 for Germany, Matano & Naticchioni, 2012 for Italy). These spatial
disparities in wages occur due to differences in skills across regions (Combes et al.,
2008) where non-routine, highly skilled-relevant tasks locate in urban areas and
401
routine and capital-intensive tasks are located in smaller regions (Scott, 2010).
However, some evidence points to that this urban premium is not universal, but
rather that it primarily pertains to workers with cognitive and people skills, or more
generally non-routine skills (Bacolod et al., 2009; Andersson et al., 2013).
One source of the urban wage premium is that larger cities offer thicker
markets for skills. Gan & Li (2004) show that the probability of a match on the job
market increases in the number of firms and worker candidates.5 In the model by
Hesley & Strange (1990), agents are heterogeneous and have imperfect information, and they choose a city knowing the number of other agents in the city
(but not their characteristics). The model predicts higher expected quality of the
match between job requirements and skills, which leads to a positive relationship
between workers wages, productivity and city size. Indeed, Duranton & Puga
(2004) maintain that one family of micro-foundations for agglomeration economies, i.e. local interactions between workers and firms that stimulate productivity, is
matching. For university graduates initiating a labour market career, these
properties of thick markets for skills are likely to be of crucial importance in their
location decision since it influences the likelihood of getting a quality job as well as
the wage.
The literature on the urban wage premium also documents a growth effect
which takes place over time. This growth effect can reflect human capital
externalities or matching efficiency over time in the sense that workers learn
what they do well and switch jobs until they find a good match between job
requirements, employer characteristics and their skill set and preferences. Topel &
Ward (1992) found that a large share of the wage increase early in the labour
market career was due to new workers changing jobs more often which is
consistent with the matching argument in labour economics, where a good match
signals higher productivity and higher wages. The uncertainty surrounding new
entrants on the labour market diminishes with time leading to stable employment.
Moreover, this finding is in line with studies on the urban wage premium that
generally find that individuals in thicker labour markets change jobs more
frequently than their more rural counterparts, which is one source of the larger
wage increases for urban workers (Wheeler, 2006; Yankow, 2006). This matching
efficiency is furthermore found to be higher for skilled workers (Matano &
Naticchiono, 2012), which may be explained by that education enhances worker
specialization. In line with this, Andersson & Thulin (2013) find that the marginal
effect of spatial employment density on the rate of inter-firm job switching is
considerably higher for better educated workers. The faster observed wage growth
of cities thus partly emanates from greater possibilities, and willingness, of skilled
workers to change jobs that improve matching processes on the labour market.
These effects are essentially linked to the internal labour market characteristics of
larger city regions.
Based on these arguments, university graduates entering the labour market in
larger cities upon graduation are expected to show higher initial wages as well as
higher wage growth associated with more frequent job switching, reflecting
advantages of thick markets for skills such as a more efficient matching processes
and early job market career development. In particular, for skilled workers such as
university graduates, a move to a big city may be considered an investment for a
long-term reward on human capital.
402
L. Ahlin et al.
3. Data, Variables and Descriptives

3.1. Description of Data
We study a cohort of Swedish graduates in the year 2000 in the social science, the
natural science and engineering.6 Our data are drawn from several databases
belonging to the Swedish population register maintained by Statistics Sweden,
which consists of the entire Swedish population aged 16 years and over. The full
set of university graduates in 2000 in all fields comprizes 37,699 individuals. We
focus on graduates with a degree in engineering, the natural or the social sciences
and who are in the age interval of 2229 years.7 We impose a number of
restrictions on the graduates in our sample, and these are detailed in the Appendix.
Our final dataset include 10,109 graduates in 2000.
The data include several characteristics of the university graduates. First, we
have information on innate attributes such as gender, parents educational level,
age, immigration status (first- or second-generation immigrant), birthplace, place of
residence since age 18 as well as information on family formation. Second, the data
include characteristics of their educational background such as high-school grades,
educational field at university, location of university studies in terms of both region
and specific university. As the data on individuals are linked to employers, we can
trace the graduates throughout the first years of their labour market career (2001
2009). In every year, we have information on employment and from this we can
identify individuals occupational codes, the sector affiliation of the employers and
the establishment size of the employers.8 By following the individuals over time,
we are able to derive measures of tenure, job changes as well as shifts in occupation
and sector between pairs of years.
Graduates with employment in a given year are defined as those with at least
100,000 SEK in total income that year. Research on Swedish data where register
data are compared with survey data shows that this is a valid definition of
employment, and that it also alleviates issues associated with lack of information on
working hours (Antelius & Bjrklund, 2000).9
3.2. Variables and Descriptive Statistics

To distinguish between different types of regions, we employ a categorization of
municipalities developed by the Swedish Board of Agriculture, in which
municipalities are categorized as belonging to a large urban region, city region or
the countryside.10 Only the Stockholm, Gothenburg and Malm regions qualify as
large urban regions. City regions, in principle, consist of medium-sized cities with
surrounding small universities. We create dummy variables for urban regions and
city regions, respectively, and use countryside as the reference group throughout
the analysis (unless otherwise stated). To measure wages and wage growth, we use
the total wage income of each graduate during a calendar year.
Table 1 presents basic descriptive statistics of two groups of graduates. The left
column refers to the full set of graduates in 2001, irrespective of whether they have
a job or not subsequent graduation. The second column reports data for a balanced
sample of those graduates with a job all years 20012009. Some notable patterns
are the majority of the graduates have a degree in engineering (45%), followed by
social sciences (40%). Most of the graduates enter the private sector (67%) and
about 51% start their labour market career with a job in a multinational enterprise
403
Table 1. Descriptive statistics for graduates in their first year on the labour market after
graduation
Full sample
Variable
Wage (in EUR)
Age
Male
Urban regions
City regions
Countryside regions
Social sciences
Natural sciences
Engineering
Parent with higher education
Average high-school grade
Married, spouse with HE
Married, spouse without HE
First-generation immigrant
Second-generation immigrant
Works in MNE
Private sector
Establishment size
Number of classmates
Parent in manager position
Percentage of graduates with a job in 2001
Urban regions
City regions
Countryside
No. of observations
Balanced sample
Mean
Std. Dev.
Mean
Std. Dev.
28,355
26.40
0.56
0.64
0.26
0.10
0.40
0.15
0.45
0.45
3.77
0.03
0.04
0.04
0.03
0.51
0.67
579.26
39.12
0.18
88.5%
89.3%
87.6%
86.1%
10,109
12,451
1.50
33 400
26.50
0.69
0.66
0.24
0.10
0.37
0.11
0.52
0.44
3.83
0.03
0.03
0.02
0.03
0.60
0.79
580.40
44.69
0.19
4,559
8835
1.41
0.76
1184.32
58.99
0.54
1229.07
62.86
Notes: The table reports basic descriptive for Swedish university graduates who graduated in 2000 and entered the
labour market in 2001. The exchange used is the average exchange rate from 8 February 2012 to 8 February 2013
(1 EUR = 8.6792 SEK).
(MNE). A large fraction of the graduates come from families with a study tradition,
as indicated by that 45% of the graduates have at least one parent with a university
background. We also observe some differences between urban, city and countryside regions in terms of the fraction of graduates that are employed the first year
after graduation.
4. Location, Wage Dynamics and Labour Market Mobility of University

Graduates: An Initial Description
4.1. Location Choice of University Graduates
The concentration of graduates to urban regions is significantly higher than what is
motivated by the general concentration of economic activity to these regions.
Table 2 reports the percentage of all workers, workers with long university
education and the balanced sample of employed graduates working in different
types of regions in 2001 and 2009, respectively. Almost two-thirds of the graduates
in our sample enter the labour market in large urban regions, and their spatial
distribution remains unchanged over the period 20012009.
404
L. Ahlin et al.
Table 2. The spatial distribution of different workers in 2001 and 2009, fractions of total
(per cent)
All workers
Large urban regions

City regions
Countryside
University educated workers
University graduates
2001
2009
2001
2009
2001
2009
39
29
32
40
30
30
51
29
20
50
30
20
66
24
10
66
24
10
Note: The table reports the percentage of all workers, all workers with a university education of at least 3 years and
our balanced sample of graduates in 2000 that work in different types of regions in 2001 and 2009, respectively.
Table 3. Migration patterns of university graduates in 2000

Work region
Home region
University region
Urban regions
City regions
Urban regions
City regions
Countryside regions
Urban regions
City regions
Countryside regions
Urban regions
City regions
Countryside regions
2,000
704
803
541
813
827
107
143
132
114
952
818
63
52
92
27
193
431
Notes: The table reports the number of university graduates in 2000 showing a specific migration pattern as judged
by the type of region they start to work in 2001. Home region is defined as the regions in which the university
graduates lived at the age of 18. There are no universities in regions classified as countryside.
The location choice and corresponding migration patterns of graduates may

take many forms (Faggian & McCann, 2009b). Table 3 reports mobility patterns
from the home region (where they lived till age 18) to the graduation region (the
region where the university is located in) as well as to the work region (the place
where their work establishment is located) of the graduates. It is most common for
university graduates to stay in the same region type throughout the period (nonmigrants). 33.5% of all university graduates are from the same type of region in
which they studied and start to work. This type of behaviour is most common
among those originating from urban regions (70%). Leaving an urban region upon
graduation is an uncommon mobility behaviour (589 out of 4,096). The big
picture is thus that a majority of the university graduates are attracted to urban
regions for work no matter where they came from previously. This suggests that
there is something that attracts them to denser regions in the first place. Urban
regions are net receivers of human capital, whereas both city regions and
countryside regions are net senders.
4.2. Wages, Wage Growth and Job Switching

Table 4 presents wage levels and wage growth for the balanced sample of
university graduates by region. The level of nominal wages is higher in urban
regions, but does not differ too much between city and countryside regions. As
expected, the average yearly wage growth is higher in urban regions.
405
Table 4. Wages in 2001 and wage growth 20012009 of university graduates in 2000 by
region type
Urban regions
City regions
Countryside regions
All 2001
All 2009
Mean yearly wage growth (%)
33,001
28,943
28,457
53,871
42,979
43,680
6.3
5.1
5.5
Note: Wage levels are in Euro. The exchange used is the average exchange rate from 8 February 2012 to 8
February 2013 (1 EUR = 8.6792 SEK).
35%
30%
25%
20%
15%
10%
5%
0%
2002
2003
Movers
2004
2005
2006
Urban regions
2007
City regions
2008
2009
Countryside regions
Figure 1. Average rate of job switching for graduates by location behaviour subsequent
graduation.
Figure 1 shows the yearly rate of job switching among four groups of graduates.
The first category is Movers and refers to graduates that move across region types
over years. This group should show the highest rate of job switching, as the
definition of this category by necessity implies that they are mobile. The remainder
three categories, i.e. urban, city and countryside regions, refer to graduates that
move to the respective type of region and remain there during the entire sample
period. For each respective category, the rate of job switching is measured as the
number of graduates that change employer as a fraction of graduates in that region
and year.
The rate of job switching is between 15% and 20% yearly in urban regions,
which is about twice as high as the rate of job switching in city and countryside
regions. This pattern is consistent with urban regions offering thicker markets for
skills and thus better opportunities for job hopping in the early job market career,
which could also explain the higher wage growth of graduates in urban regions.
Corresponding graphs for men and women reveal similar patterns.11
5. Identification Strategy and Results

5.1. Identification Strategy
The main aim of the subsequent analyses is to unravel whether the observed
advantages of urban regions in terms of employment probability, initial wages and
wage growth reflect an effect of the urban environment, i.e. thicker market for
skills, or if it is simply an artefact of that more able graduates self-select to urban
regions. To identify the effect of market thickness, we estimate the influence that
406
L. Ahlin et al.
entering urban labour markets has on wage levels and wage growth, while
controlling for several confounding factors.
We undertake a three-pronged analysis. First, we assess defining characteristics
of graduates that locate in urban regions. Second, we analyse the influence of
location on employment probabilities and initial wage levels. Third, we estimate
the influence of location on wage growth. We describe our strategy to identify the
effect of location in each set of analysis in the following sub-sections.12
5.1.1. Selection of university graduates to urban regions. Our main hypothesis is that
urban regions are attractive primarily because of job- and career-related advantages
associated with thick markets for skills. This suggests that university graduates
moving to urban regions to enter the job market may be a selected group, with
greater ambitions and motivations. To test this, we estimate a Probit model with
observable individual characteristics as regressors:
PrUi 1jxi Ux0i C
x0i C a I0i b
uE
h h h
X
r
1a
dr Hr ei
1b
The dependent variable, Ui, is a dummy variable which is 1 if university

graduate i starts to work in an urban region 2001, i.e. the first year on the labour
market after graduation, and 0 otherwise. The matrix I includes individual
characteristics reflecting motivation, socio-economic background and marital
status. We use high-school grades as a proxy for ability or ambition (cf. Grogger
& Eide, 1995; Miller 1998; Geiser & Santelices, 2007). A large literature shows that
workers in urban regions are indeed a select group, such that spatial sorting on
(unobserved) abilities explains a large fraction of the general urban wage premium
(Combes et al., 2008). High-school grades are one observable individual
characteristic related to ability and ambition. Socio-economic background is
reflected by a dummy variable which is 1 if any of the parents have a long
university education. A general expectation is that individuals with better educated
parents are more prone to move to urban regions to gain reward to their human
capital, for instance due to support and encouragement from the home
environment (cf. Dustmann, 2004). We also include indicators of whether the
individual is married and whether the partner has a long university education. The
latter is intended to reflect the dual labour market career problem, where university
graduates with highly educated partners are expected to be more prone to move to
larger regions (Costa & Kahn, 2000). Descriptive statistics for all these variables are
presented in Table 1. Furthermore, we include age and gender. E and H represent
dummies for the educational specialization (or field of education) of individual i
and the higher educational institution at which individual i studied.
5.1.2. The influence of location on employment probabilities and initial wage levels. The
matching advantages associated with thicker markets for skills suggest that the
probability of finding a job as well as the initial wages of university graduates
entering the labour market in urban regions should be greater (Hesley & Strange,
1990; Gan & Li, 2004). To test this, we employ a two-step selection model
(Heckman, 1979). First, we estimate a Probit model for employment, where the
main interest is on the influence that a location in an urban region has on the
probability of being employed. To make sure that the estimated influence of
407
location is simply not an artefact of that better graduates self-select to bigger

cities, we control for a full set of graduate characteristics. Second, we estimate a
wage equation for initial wages, in which we control for employment selection by
including the Inverted Mills Ratio from the Probit model in stage one.13 Again,
the main interest is on the estimated influence of living in an urban region. The
wage equation is given as follows:
In wi a /1 DiU /2 DiCR I0i b X0i h xki ei
where wi is the wage of graduate i in 2001, i.e. the first year in which graduates
in 2000 are observed on the labour market. DiU and DiCR are dummies for living in
an urban region and city region, respectively. The reference group is the
countryside. The I matrix includes controls aimed at capturing the influence that
innate individual characteristics has on wages. These are age, dummy for first- and
second-generation immigrant, respectively, gender, high-school grades and a
dummy for whether at least one parent has a long university education and
whether the graduate is married, respectively. We also include an indicator of the
field of study. represents characteristics associated with the work and includes
employer size in terms of employees, two-digit Nomenclature statistique des
Activits conomiques dans la Communaut Europenne (NACE) sectors dummies, occupation dummies at the one-digit ISCO-88 classification system as well as
a dummy for working in the private sector.14 is the selection adjustment term,
i.e. Mills estimated in the first-stage Probit model, and accounts for the possible
bias stemming from the employment selection process.15
A key issue in the Heckman selection model concerns the so-called exclusion
restrictions, which are needed to properly identify the selection process (Puhani,
2000). In our empirical context, this means that we need variables that influence
the employment probability (selection equation) but not the wage level (outcome
equation), and such variables are in principle instruments. We follow the intuition
of Borjas et al. (1992) and other studies in labour economics, and consider the
wage to primarily depend on intrinsic individual attributes. Our main instruments
are instead aimed at reflecting social network characteristics of the graduates that
should influence information about job opportunities, but not the wage level
directly. First, we include the number of university classmates that are employed and live
in the same region as the graduate. University classmates are defined as people in the
same cohort with regard to university and years as well as subject of study. This
variable is intended to capture the effect peers and social networks have on the
probability of getting a job. The well-known work by Granovetter (1995)
illustrated the important role played by personal networks and friendships in
acquisition of information about job opportunities. Recent work by Marmaros &
Sacerdote (2006) also shows that strong friendships and networks indeed tend to
form between university classmates.16 We argue that the number of employed
university classmates in the region in which a graduate lives is an indicator of the
extent of a graduates social network in that region that can help in finding a
suitable job. These networks are yet less likely to be important in explaining the
initial wage, as the wage is more associated with the intrinsic individual attributes.
Second, we use an indicator whether the graduate has at least one parent in a
management position in industry. The argument is that having a parent with a higher
position in industry implies that the parent has stronger connections and networks
in industry that could help the graduates chances of getting a first job. In addition
408
L. Ahlin et al.
to these two main instruments, we also include university dummies as instruments.

Although some studies show that there are indeed university effects in earnings
(Betts et al., 2013), previous research using micro-data on Swedish graduates finds
little or no evidence for such effects once one controls for basic ability indicators
and location as we do in this study (see e.g. Eliasson 2006; Holmlund, 2009).
There are still various ways in which the likelihood of getting a job may be
influenced by the university at which one studies. Employers may, for example,
use the university background of graduates as a signalling device in scanning
job applications. Another argument is that there may be heterogeneity across
universities in terms of networks to the business sector and how they work with
alumni to help graduates get a job.
To probe the robustness of the selection model on initial wages, we also exploit
the panel structure of the data, and estimate the model in (2), excluding the
selection adjustment term, on a panel dataset allowing us to account for worker
time-invariant unobserved heterogeneity of graduates.17 The motivation for this is
that a large literature shows that unobserved time-invariant individual skills have a
large influence on estimates of the influence of location on wage, as individuals
may sort themselves spatially on such unobserved factors (Combes et al., 2008;
Andersson et al. 2013).
5.1.3. Wage growth. To test the effect of location on subsequent wage growth,
we focus on a balanced panel of graduates that remain employed all years during
the period 20012009 and estimate the following model:

wit1
^U / D
^ CR / D
^ CSR I0 b X0 h xk
ln
a q In wit /1D
i
2 i
3 i
i
it
it
wit
eit
This is essentially the same model as in (2), but with the yearly wage growth as the
dependent variable and the wage level as an additional independent variable. i is a
selection adjustment term that we calculate from a Probit model explaining the
probability that a graduate belongs to the set of graduates that are employed all years.18
As indicated by the -sign, the dummies for the different types of regions are
^U, D
^ CR and D
^ CSR denote stayer in
defined somewhat differently than before. D
i
i
i
urban, city and countryside regions, respectively. They are 1 for those university
graduates that move to the type of region in question and remain there all 9 years
they are observed. This means that the dummy for urban regions is 1 for university
graduates that start to work there in 2001 and continue to work in urban regions
all subsequent years. The reference group is movers, i.e. university graduates that
move across region types over years. The reason we do this distinction follows
directly from theory.
Theory suggests that working in thicker markets increases wage growth because
job switching is cheaper and access to a variety of qualified job positions is better
(cf. Finney & Kohlhase, 2008; Andersson & Thulin, 2013). Search processes
towards a quality match in which graduates also learn about their own job
preferences and what they do well should therefore be quicker (Wheeler, 2006).
To reap these thick market advantages, however, workers should remain in dense
environment during their labour market career.19
We define four categories of workers: those that remain in either urban
regions, city regions or the countryside and a fourth category comprising those
409
workers that move across region types over years. By construction, these
categorical variables are time-invariant, which means that a panel estimator with
worker fixed effects is unfeasible. A simple ordinary least squares (OLS) estimator
with extensive controls and a sample selection adjustment term accounting for that
graduates employed all years may be a select group is our best option. At the same
time, our simple model includes a rich set of time-invariant individual attributes
that are typically unobserved in many other dataset. We, for example, control for
high-school grades, which in the literature on the returns to education is often used
to proxy latent ability (cf. Eliasson, 2006). As seen below, we also undertake sample
splits to make the samples more homogenous.20
5.1.4. Sample splits. For all models, we present results for all university graduates
as well as for those that grew up at countryside and did not study at a university
located in urban regions. The latter group of graduates are henceforth referred to as
active movers because those who live and work in urban regions in this group have
made an active move. In the full sample of all university graduates, we mix active
movers and graduates that grew up in the urban regions, studied there and simply
remained there after graduation. In this sense, they made no major location choice.
5.2. Results
5.2.1. Who starts working in urban regions? Table 5 reports the relationship
between individual innate characteristics and the likelihood of starting the
professional career in a large urban region. In general, the results confirm that
those graduates starting to work in urban regions are a select group. Having highly
Table 5. Selection of graduates to urban regions. Probit regression, dependent variable:
1 = first job in a large urban region

High-school grade: in logarithms
Age (ln)
Male
Natural sciences
Engineering
Constant
No. of observations
All
Active movers
0.262***
(0.0737)
0.352***
(0.0701)
0.225***
(0.0286)
1.233***
(0.101)
1.917***
(0.267)
0.138***
(0.0303)
0.179***
(0.0436)
0.115***
(0.0325)
7.619***
(0.908)
8,755
0.174
(0.158)
0.646***
(0.169)
0.153**
(0.0601)
0.111
(0.210)
1.183**
(0.521)
0.0613
(0.0632)
0.0878
(0.0840)
0.295***
(0.0679)
3.873**
(1.771)
2,054
Note: The table reports coefficients of the independent variables. Robust standard errors in parentheses. Active
movers refer to a restricted sample of university graduates that originate from the countryside and did not study at a
university in an urban region.
*, ** and *** denote statistical significance at the 10-, 5- and 1-percentage level, respectively.
410
L. Ahlin et al.
educated parents, being older and having studied social sciences are all significantly
correlated with the likelihood of starting to work in an urban region. In the full
sample, high-school grades are also positive and significant. Furthermore, having a
spouse without higher education is negative and significant, whereas having a
spouse with higher education is negative but insignificant for active movers only.
One explanation for this is that a spouse complicates migration, because the
migration decision may transform from an individual to a household choice and
this is aggravated when the spouse is not highly educated. Among those graduates
that grew up in the countryside and studied outside urban regions, we find no
statistically significant difference between men and women in terms of their
inclination to start working in an urban region upon graduation. We conclude that
graduates moving to urban regions are a select group. They appear to be more
motivated as judged by their high-school grades and the educational background of
their parents.
5.2.2. The Initial Urban Wage Premium. Table 6 reports estimates of the variables
in the selection as well as outcome equation of the selection model for initial
wages. Starting with the determinants of the probability of being employed, we see
as expected that living in an urban region has a positive and significant effect on the
likelihood of being employed. The type of region a graduate lives in thus matter
over and above individual attributes. Most of the individual attributes also have the
expected sign and are significant.21 Older graduates with better high-school grades
and a degree in engineering are more likely to be employed. This holds for
graduates in general as well as for the subset of active movers, although high-school
grades are only significant at the 10% level and having a degree in social sciences is
insignificant for the latter group. First-generation immigrants have lower probability of and males somewhat higher probability of being employed when the full
sample of graduates is considered. For the subset of active movers, however, we
find no effect of any of these variables. This may be explained by that those coming
from the countryside and study at universities may be a more strongly selected
group, such that gender and immigrant differences are reduced. Being married
(spouse with or without higher education) has no significant effect on the
probability of being employed.
Turning to the instruments, we find that the number of university classmates in
the residential region that have a job is significant and positive for both samples.
This is in line with expectations and supports the hypothesis of the role played by
local friendships and social networks in the process of finding a job. Many of the
(unreported) university dummies that we also use as exclusion restriction are still
significant, which may be explained by any of the two reasons discussed in the
previous section. We find no effect of having at least one parent with a
management position in industry on the probability of getting a job.
The results in the outcome equations show that the urban initial wage
premium is highly significant for both groups of graduates. The point estimate
suggests a 5.7% wage premium for those graduates starting to work in urban
regions.22 Such an urban initial wage premium is consistent with the predictions
from matching models, which suggest that the expected quality of a job match
increases with city size (Hesley & Strange, 1990). There is no evidence of a wage
premium for graduates in city regions, suggesting that such effects require a critical
mass and are a truly urban phenomenon.
411
Table 6. Initial wage in 2001 of university graduates. Heckman two-step regression (Probit
+OLS)
Urban region (residential)

City region (residential)
High-school grade: in logarithms
Age (ln)
Male
Natural sciences
Engineering
Classmates in residential region
with a job
Private sector
All
All
Active movers
Active movers
Selection
equation
Outcome
equation
Selection
equation
Outcome
equation
0.204***
(0.0600)
0.0749
(0.0618)
0.0741
(0.101)
0.0762
(0.0940)
0.584***
(0.139)
2.774***
(0.347)
0.0670*
(0.0380)
0.272***
(0.0893)
0.0845
(0.100)
0.226***
(0.0497)
0.187***
(0.0527)
0.0332
(0.0464)
0.00338***
(0.000438)
0.0567***
(0.00994)
0.00798
(0.0103)
0.0101
(0.0146)
0.00799
(0.0140)
0.176***
(0.0204)
0.625***
(0.0590)
0.0696***
(0.00603)
0.0339**
(0.0159)
0.00735
(0.0164)
0.0137
(0.00999)
0.0518***
(0.00745)
0.410***
(0.101)
0.245***
(0.0933)
0.267
(0.248)
0.189
(0.189)
0.523*
(0.311)
2.757***
(0.735)
0.0384
(0.0851)
0.0673
(0.267)
0.0548
(0.299)
0.105
(0.110)
0.147
(0.0974)
0.0435
(0.108)
0.00490**
(0.00241)
0.0576***
(0.0169)
0.00113
(0.0149)
0.0570*
(0.0304)
0.00769
(0.0291)
0.118***
(0.0419)
0.675***
(0.111)
0.0721***
(0.0124)
0.0215
(0.0365)
0.0414
(0.0446)
0.0132
(0.0182)
0.0252*
(0.0152)
Establishment size (ln)

Mills
Constant
8.656***
(1.195)
8964
0.137***
(0.0108)
0.0150***
(0.00174)
0.152***
(0.0349)
5.284***
(0.312)
7842
8.466***
(2.536)
2132
0.0863***
(0.0205)
0.00788**
(0.00349)
0.226***
(0.0619)
4.976***
(0.398)
1836
Notes: The selection equation includes dummy variables for university, and the outcome equation dummy
variables for occupation and industry. Robust standard errors in parentheses. Active movers refer to a restricted
sample of university graduates who originate from the countryside and did not study at a university in an urban
region.
Individual attributes are also significantly associated with the wage outcome.
High-school grades are positive and significant with an elasticity of 18% for the
sample of all graduates. This means that 10% higher average high-school grades are
associated with a 1.8% higher initial wage. This is consistent with the argument
that high-school grades proxy underlying motivation and ambition that may
influence initial wages. We also find a difference in the general wage level between
males and females, where the former group has approximately 7% higher initial
wages on average. We find no effect of foreign background. Having a spouse (with
412
L. Ahlin et al.
or without higher education) is not significant for the probability of getting or for
initial wage. Age and working in the private sector are also positively associated
with initial wages. The selection adjustment term (Mills ) is significant in both
outcome equations, indicating the statistical importance of accounting for the
employment selection process.
As a robustness test of the main results in Table 6, Table 7 reports panel
estimates with worker fixed effects. Consistent with a large literature (cf. Combes
et al., 2008), the estimated urban wage premium falls when accounting for timeinvariant unobserved abilities. The reported estimates of the premium in Table 7
are in the interval of 2.2% and 4%, the latter applying to graduates from the
countryside. Though these estimates are still significant but lower, it is important to
recognize that they answer a somewhat different question. The estimates in Table 6
pertain to initial wages, i.e. the first years after graduation. The location dummy
variables are here instead based on graduates that move to an urban region in years
after 2001 (i.e. the 20022009 period). We still find significant thick market effects
that appear to be confined to the larger urban regions. The negative effect of
tenure can be explained by a strong correlation with job switching. It is on average
better to switch jobs often at the beginning of a career instead of remaining with
the same employer. This is consistent with the matching hypothesis. We also find
Table 7. Level effects on wages-dependent variable: wage level (Panel FE estimator)
Urban region (work)

City region (work)
Age (ln)
Private sector
Tenure
Tenure squared
Job change
MNE
Constant
No. of observations
R2
No. of individuals
All
Active movers
0.0224**
(0.0110)
0.00452
(0.0103)
0.0590***
(0.00722)
0.0624***
(0.00894)
1.811***
(0.0305)
0.0215*
(0.0112)
0.0190***
(0.00312)
0.00118***
(0.000432)
0.00347
(0.00497)
0.00429**
(0.00194)
0.0311***
(0.00648)
1.865***
(0.118)
41,031
0.287
4,559
0.0389*
(0.0206)
0.00777
(0.0162)
0.0699***
(0.0142)
0.0832***
(0.0174)
1.815***
(0.0588)
0.0239
(0.0226)
0.0207***
(0.00610)
0.00154*
(0.000838)
0.00614
(0.00980)
0.00353
(0.00409)
0.0215*
(0.0128)
1.785***
(0.265)
9,594
0.309
1,066
Note: All regressions include dummy variables for year, university, occupation and sector. Robust standard errors in
parentheses. Active movers refer to a restricted sample of university graduates who originate from the countryside
and did not study at a university in an urban region.
413
that switching to the private sector or to a multinational corporation raises wages

for the sample of all graduates. A further result is that becoming married is
negatively associated with the wage.
5.2.3. Wage-growth premium. Table 8 shows the estimated wage growth effects
for those who remain in the same type of regional category throughout the entire
period. In these estimations, the underlying data are a balanced sample of 4,559
Table 8. Wage growth of university graduates by location. Heckman two-step regression
(Probit + OLS)
Urban region (work) all years

City region (work) all years
Countryside region (work) all years
Lagged income (ln)
Male
High-school grade (ln)
Natural sciences
Engineering
Age (ln)
Private sector
Tenure
Tenure squared
Job change
MNE
Mills
Constant
No. of observations
R2
All (outcome)
Active movers (outcome)
0.0221***
(0.00365)
0.00745
(0.00462)
0.00903
(0.00745)
0.391***
(0.00457)
0.00830
(0.00848)
0.0169**
(0.00852)
0.00830
(0.00813)
0.0556***
(0.0114)
0.0267***
(0.00518)
0.0347***
(0.00436)
0.102***
(0.0311)
0.0303***
(0.00645)
0.0140***
(0.00272)
0.00152***
(0.000351)
0.0254***
(0.00512)
0.00606***
(0.000941)
0.0137***
(0.00371)
0.169***
(0.0180)
2.714***
(0.146)
36 472
0.181
0.0355***
(0.00747)
0.00282
(0.00739)
0.0146
(0.00989)
0.429***
(0.00972)
0.0534**
(0.0209)
0.0102
(0.0204)
0.0256
(0.0206)
0.0779***
(0.0213)
0.0124
(0.00981)
0.0176**
(0.00773)
0.0632
(0.0574)
0.0399***
(0.0123)
0.0169***
(0.00519)
0.00189***
(0.000667)
0.0243**
(0.00983)
0.00345*
(0.00189)
0.0134*
(0.00722)
0.281***
(0.0395)
3.222***
(0.270)
8528
0.204
Notes: The selection equation includes dummy variables for university, and the outcome equation dummy
variables for occupation and industry. Robust standard errors in parenthesis. Active movers refer to a restricted
sample of university graduates who originate from the countryside and did not study at a university in an urban
region. For the urban, city and countryside region stayer dummies, the reference group is movers.
414
L. Ahlin et al.
graduates that are employed all years during the 20012009 period.23 Graduates
that move across region types over the 9 years on the labour market are the
reference group in all specifications. The estimations include a selection adjustment
term (Mills ) obtained from a Probit estimation of the likelihood of being
employed all years 20012009.24
We find a strong wage growth effect for those graduates operating in urban
regions all years. The point estimates imply an annual wage growth premium of 2
4%. These findings are consistent with the hypothesis that the thicker labour
markets for skills in urban regions imply better long-term reward to human capital
and a stronger early labour market career development.25 Looking at the controls,
high-school grade has a significant positive effect on wage growth, with an
elasticity in the order of 58%. This is in line with the findings of Miller (1998),
who shows that high-school grades are able to predict long-term earnings. Private
sector workers experience 34 percentage points faster wage growth per year. We
also find that switching jobs provides an additional boost to the wage growth by
approximately 2 percentage points. Working in larger establishments has a positive
impact on the annual wage growth, although the effect appears small. The
estimates suggest that doubling the number of workers in an establishment
increases wage growth by about half a percentage point. Tenure has a significantly
negative influence on wage growth, although the negative effect falls as tenure
increases. Mills is significant in both models, again confirming the importance of
accounting for the underlying employment selection process.
In summary, the results reported in Tables 58 provide strong support for the
hypothesis that larger urban regions provide advantages associated with market
thickness for skills. Working in a dense environment with a thick labour market with
many employers to choose from facilitates for workers to search for a good match
between their skills and aptitudes and the requirements demanded by the job. The
search process involves frequent job switching in the beginning of the professional
career which of course provides a disadvantage for less dense regions that lack the
range of employers found in more densely populated regions. The evidence
provided by the regressions also highlights the general role job switching has for
wages. Changing employer, regardless of region, yields a positive wage effect which
can be interpreted as a result of a continuous job match improvement. We also show
that these results hold for graduates in general as well as for workers moving to urban
regions upon graduation from university with no prior attachment to such regions.
6. Conclusions
The vast majority of graduates move to large cities to begin their professional
labour market careers. We set out to understand this spatial concentration of
university graduates. This is an issue of importance, not least since the spatial
distribution of highly educated workers is considered a key determinant of
divergence in growth, incomes and well-being between regions (Moretti, 2012).
Our main finding is that the attraction of urban regions among university
graduates is to be found in their labour market characteristic, i.e. they provide
thicker markets for skills. This is manifested not only by a higher probability of
being employed and higher initial wages but also by faster wage growth and more
frequent job switching during the early labour market career of university graduates
moving to urban regions. Our analyses of the probability of being employed
415
further emphasize the role friendships and social networks developed with
classmates during university studies play in getting a first job.
Urban regions are net receivers while all other areas are net senders of
graduates. The graduates who start to work in large urban regions are not a random
sample, but rather a select group of individuals with better grades and more
educated parents i.e. graduates for which labour market characteristic and career
prospects are likely to be more important.
We find a robust and positive wage premium from entering the thicker labour
markets in urban regions. University graduates who start their professional career in
large urban regions earn around 56% higher initial wages as compared with
graduates working in other types of regions. Panel regressions with fixed
individual-specific effects reveal that moving to an urban region during the first
years on the labour market tend to give an immediate nominal wage increase by
approximately 24%. These results are consistent with models emphasizing that
expected match qualities rise with market thickness.
We also show that graduates who choose to begin their professional labour
market career in urban regions experience higher wage growth, even after
controlling for employment selection. In fact, those graduates entering the labour
market in an urban region directly after graduation and remain there during the
eight first years on the labour market show the highest average yearly wage
growth. One reason for this is that these graduates enjoy benefits of agglomeration
economies, i.e. interactions between workers and their local environment that
lead to productivity gains. A source of such agglomeration economies is matching,
where workers benefit from the thicker labour markets in urban regions. We
indeed find higher rates of job switching among university graduates operating in
urban regions, consistent with the idea that graduates in denser regions are in a
better position to shop around in the early job market career. These results are
perfectly in line with Wheelers (2006) findings for young workers in the USA: an
important aspect of learning in cities may involve individuals learning about what
they do well (p. 162).
A further finding is that the described effects of market thickness for skills
seem to be confined to big urban cities. We distinguished between three types of
regionsurban, city and countryside regionsbut we found no statistically
significant difference between city and countryside regions in terms of neither
wage levels nor wage growth. This suggests that a critical mass in terms of
population density is needed to reap the advantages of market thickness.
During the last decades, there has been a debate in the literature as regards the
relative roles of jobs on the one hand and amenities in the other in attracting human
capital (see e.g. Storper & Scott, 2009; Niedomysl & Hansen, 2010; Scott, 2010).
Our results for the case of graduates lend support to that labour market characteristics,
and in particular early labour market career prospect, are key attractors.
Notes
1. The concentration of university graduates to larger urban regions is not unique for Sweden. Costa & Kahn
(2000) report an increasing concentration of human capital to large metropolitan areas in the USA over the
period 19401990.
2. A main reason for this is that larger cities tend to host a disproportionately large fraction of headquarter
functions of industry (Bel & Fageda, 2008), knowledge-intensive business services (Jacobs et al., 2013) as well
as governmental agencies focusing on, e.g. qualified investigational work.
416
L. Ahlin et al.
3. Such a search process is also, in part, a process in which the workers learn about what they do well (cf.
Wheeler, 2006).
4. Although these authors give plausible reasons for this omission (lack of suitable data and potentially different
prospects due to child-bearing), they still omit a large fraction of the population as well as confound the
micro-foundations for the observed decision if the location-decisions of some skilled males are entwined with
the decisions of their spouses.
5. To be precise, they show that a market with five job openings and five candidates have much lower matching
probability than a market with 50 openings and 50 candidates.
6. Graduates are defined as those who have applied for and been granted a degree. One caveat of this approach is
that we cannot study those that drop out or get a job prior to having finished their final thesis.
7. We thus exclude subject fields associated with jobs such as artists, teachers, medical doctors and armed forces,
i.e. jobs for which labour markets in Sweden are rather particular. The location of employers as well as
employment decisions and wage setting are strongly regulated by state-regulations in these types of jobs.
8. The workplace that the individual is allocated to during a year is determined by the workers workplace in the
month of November each year. This may understate our measures of inter-firm (or inter-workplace) mobility
as we cannot observe the employer in other months (Andersson & Thulin, 2013).
9. Antelius & Bjrklund (2000, p. 354) state: In order to reduce the impact of working time on the estimates,
we eliminated those observations with low annual earnings from the sample. When this restriction was set to
SEK 100,000 the estimates were quite close to those obtained using hourly earnings.
10. Several different criteria are used to categorize municipalities, including in- and out-commuting as a fraction
of total employment, percentage of populated surface and distance to large city (Swedish Agricultural Board,
Report 2009, 2).
11. There is a notable decline in the rate of job switching in end of the reporting period. There are two main
explanations for this: (1) a large literature shows that labour mobility tend to be pro-cyclical and the later years
in the reporting period corresponds to the financial crisis and (2) it is natural that the average rate of job
switching of graduates declines with their time on the labour market in general, as the likelihood that they
find a good match increases over time.
12. Table A1 in Appendix defines all variables appearing in the regression analyses.
13. The wage equation is based only on those graduates who are employed in the first year after graduation.
14. The variables related to the job of the graduates only enter in the outcome equations. They cannot be
included in the selection equation because they are fundamentally related to the outcome of the selection
process. We are well aware that including variables in the outcome equation that are not in the selection
equation is a practice that is debated, as it effectively implies that we impose restrictions on the selection
process. In our empirical context, we are still primarily interested in the influence a location in an urban
region has on initial wages through matching effects associated with market thickness. We therefore argue that
it is warranted to include controls for occupations, sectors and firm size in the outcome equation to make sure
that the estimated influence of an urban location not only captures very basic differences between regions in
terms of industry compositions. We have also verified that our results are robust to the exclusion of the jobrelated variables that are not present in the selection equation.
15. Technically, Mills is defined as = bX=UbX, where (.) is the density function of the standard normal
distribution and (.) is the cumulative normal distribution coming from the first-stage model estimated through a Probit.
16. They find, for example, that placing two students in the same entering class (cohort) has a 6 effect on the
frequency of their interacting (Antelius & Bjrklund, p. 81). Clearly, friendships and social networks form
from such interactions.
17. We observe the university graduates over a period of 9 years. The within-transformation of the fixed effects
estimator means that we no longer focus on initial wages, but rather the wage-level effect associated with
moving to an urban region during the years after 2001. Parameters are identified from changes in the variables
for the same individual over time, which means that parameters for time-invariant variables cannot be
estimated. The benefit of this estimator is that we can account for unobserved time-invariant heterogeneity
not captured by the observable individual attributes, and provides a test of the robustness of the results
obtained with the simpler OLS model.
18. We use the same set of variables and instruments as in the previous selection model.
19. Moving away from an urban region to the countryside, for instance, would mean that the thick market effects
associated with matching are in all essence lost or reduced.
20. While our described strategy certainly alleviates issues of selection of good graduates to urban regions, which
may explain part of the urban premium with regard to initial wages and wage growth, we cannot rule out the
occurrence of an influence of remaining unobserved innate characteristics.
21. We choose to exclude the dummy for whether at least one parent has a long university education. The reason
is a very strong correlation between this variable and other regressors, especially the university dummies in the
selection equation.
417
22. Note that starting the work may imply different mobility patterns between the full sample and the sample of
active movers. In the former, starting to work in an urban region may simply mean that they stay where
they studied. In the latter, it means that the graduates moved to urban regions on subsequent graduation.
23. Main reasons for the drop in the number of graduates when switching to a balanced sample is that many
graduates fall out of the employment definition when having children and going abroad.
24. The selection equation is here specified in the same way as in Table 6. We find similar results as in Table 6
with regard to the influence of individual innate attributes and the exclusion restrictions on the probability of
being employed all years. The main difference is that having parents with a management position in industry
is significant and positive in the selection equation for employment all years. This supports the hypothesis that
such parents have stronger connections and networks in industry that could help their childrens chances of
employment. Graduates living in urban regions all years are also more likely to be employed all years. These
results are available from the authors upon request.
25. We have also estimated a fixed effects panel model with annual wage growth as the dependent variable,
which allow us to account for the role of time-invariant unobserved abilities in wage growth. Although this
model does not allow us to identify the effect of remaining in the same region over all nine years, we find that
switching to a job in a large urban region results in a significantly positive wage effect. As before, however, no
significant wage effect is found for city regions. These results are available from the authors upon request.
References
4Potentials (2012) Att behlla och utveckla talanger en studie genomfrd hsten 2011 i talangntverket 4Potentials [To
attract and keep talents a study conducted in the fall 2011 in the network of talents 4 potentials], Report at
4Potentials.
Andersson, M., Klaesson, J. & Larsson, J. P. (2013) The sources of the urban wage premium by worker skills
spatial sorting or agglomeration economies? Papers in Regional Science. doi:10.1111/pirs.12025
Andersson, M. & Lf, H. (2012) Small business innovation firm-level evidence from Sweden, Journal of
Technology Transfer, 37, 732754.
Andersson, M. & Thulin, P. (2013) Does spatial employment density spur inter-firm job switching? Annals of
Regional Science, 5, 245272.
Antelius, J. & Bjrklund, A. (2000) How reliable are register data for studies of the return on schooling? an
examination of Swedish data, Scandinavian Journal of Educational Research, 44, 341355.
Bacolod, M., Blum, B. S. & Strange, W. (2009) Skills in the city, Journal of Urban Economics, 65, 136153.
Badinger, H. & Tondl, G. (2003) Trade, human capital and innovation: the engines of European regional growth
in the 1990s, in: B. Fingleton (ed) European Regional Growth, Series: Advances in Spatial Science, pp. 215240,
Heidelberg, Springer.
Bel, G. & Fageda, X. (2008) Getting there fast: globalization, intercontinental flights and the location of
headquarters, Journal of Economic Geography, 8, 471495.
Betts, J., Ferrall, C. & Finnie, R. (2013) The role of university characteristics in determining post-graduation
outcomes: panel evidence from three Canadian cohorts, Journal Canadian Public Policy, 39, 81106.
Borjas, G. J., Bronars, S. G. & Trejo, S. J. (1992) Assimilation and the earnings of young internal migrants, Review
of Economics and Statistics, 74, 170175.
Bradley, S. & Taylor, J. (1996) Human capital formation and local economic performance, Regional Studies, 30,
114.
Cheshire, P. C. & Magrini, S. (2000) Endogenous processes in European regional growth: implications for
convergence and policy, Growth and Change, 32, 45579.
Combes, P., Duranton, G. & Gobillon, L. (2008) Spatial wage disparities: sorting matters! Journal of Urban
Economics, 63, 723742.
Costa, D. & Kahn, M. (2000) Power couples: changes in the locational choice of the college educated 19401990,
Quarterly Journal of Economics, 115, 12871315.
Duranton, G. & Puga, D. (2004) Micro-foundations of urban agglomeration economies, in: J. V. Henderson &
J. F. Thisse (eds) Handbook of Regional and Urban Economics, Chapter 48, pp. 20632117, Amsterdam, Elsevier.
Dustmann, C. (2004) Parental background, secondary school track choice and wages, Oxford Economic Papers, 56,
209230.
Eliasson, K. (2006) The role of ability in estimating the returns to college choice new Swedish evidence, Ume
Economic Studies No. 691, Ume, Ume University.
Faggian, A. & McCann, P. (2006) Human capital flows and regional knowledge assets: a simultaneous equation
approach, Oxford Economic Papers, 58, 475500.
Faggian, A. & McCann, P. (2009a) Human capital, graduate migration and innovation in British regions,
Cambridge Journal of Economics, 33, 317333.
418
L. Ahlin et al.
Faggian, A. & McCann, P. (2009b) Universities, agglomerations and graduate human capital mobility, Tijdschrift
voor Economische en Sociale Geografie [Journal of Economic and Social Geography], 100, 210223.
Farber, H. S. (1994) The analysis of inter-firm worker mobility, Journal of Labor Economics, 12, 554593.
Finney, M. M. & Kohlhase, J. (2008) The effect of urbanization on labor turnover, Journal of Regional Science, 48,
311328.
Gan, L. & Li, Q. (2004) Efficiency of thin and thick markets, NBER Working paper 10815.
Geiser, S. & Santelices, M. V. (2007) Validity of high school grades in predicting student success beyond the
freshman year, Research and Occasional Paper Series CSHE 6.07, Berkley, University of California.
Glaeser, E. & Gottlieb, P. D. (2006) Urban resurgence and the consumer city, Urban studies, 43, 12751299.
Glaeser, E. & Mar, D. (2001) Cities and skills, Journal of Labor Economics, 19, 316342.
Glaeser, E., Scheinkman, J. & Schleifer, A. (1995) Economic growth in a cross-section of cities, Journal of Monetary
Economics, 36, 117143.
Granovetter, M. S. (1995) Getting a Job: A Study of Contacts and Careers, 2nd ed. Chicago, IL, University of
Chicago Press.
Grogger, J. & Eide, E. (1995) Changes in college skills and the rise in the college wage premium, Journal of Human
Resources, 30, 280310.
Heckman, J. (1979) Selection bias as a specification error, Econometrica, 47, 153161.
Hesley, R. & Strange, W. (1990) Matching and agglomeration economies in a system of cities, Regional Science and
Urban Economics, 20, 189212.
Holmlund, L. (2009) The effect of college quality on earnings evidence from Sweden, Ume Economic Studies
No. 781, Ume, Ume University.
Jacobs, W., Koster, H. & van Oort, F. (2013) Co-agglomeration of knowledge-intensive business services and
multinational enterprises, Journal of Economic Geography, forthcoming.
Kim, S. (1990) Labor heterogeneity, wage bargaining and agglomeration economies, Journal of Urban Economics,
28, 160177.
Lehmer, F. & Mller, J. (2010) Interrelations between the urban wage premium and firm-size wage differentials: a
microdata cohort analysis for Germany, Annals of Regional Science, 45, 3153.
Marmaros, D. & Sacerdote, B. (2006) How do friendship form? Quarterly Journal of Economics, 121, 79119.
Matano, A. & Naticchioni, P. (2012) Wage distribution and the spatial sorting of workers, Journal of Economic
Geography, 12, 379408.
Miller, S. R. (1998) Shortcut: high school grades as a signal of human capital, Educational Evaluation Policy Analysis,
20, 299311.
Moretti, E. (2012) The New Geography of Jobs, New York, NY, Harcourt Publishing Company.
Moretti, E. & Thulin, P. (2013) Local multipliers and human capital in the United States and Sweden, Industrial
and Corporate Change, 22, 339362.
Niedomysl, T. & Hansen, H. (2010) What matters more for the decision to migrate? jobs versus amenities,
Environment and Planning A, 42, 16361649.
Puhani, P. A. (2000) The Heckman correction for sample selection and its critique, Journal of Economic Surveys, 14,
5368.
Scott, A. J. (2010) Jobs or amenities? destination choices of migrant engineers in the USA, Papers in Regional
Science, 89, 4363.
Shapiro, J. (2005) Smart cities: quality of life, productivity, and the growth effects of human capital, NBER
Working Papers 11615.
Simon, C. (1998) Human capital and metropolitan employment growth, Journal of Urban Economics, 43, 223243.
Storper, M. & Scott, A. J. (2009) Rethinking human capital, creativity and urban growth, Journal of Economic
Geography, 9, 147167.
Topel, R. H. & Ward, M. P. (1992) Job mobility and the careers of young men, Quarterly Journal of Economics,
107, 439479.
Venhorst, V., van Dijk, J. & van Wissen, L. (2010) Do the best graduates leave the peripheral areas of the
Netherlands? Tijdschrift voor Economische en Sociale Geografie [Journal of Economic and Social Geography] 101,
521537.
Venhorst, V. A., van Dijk, J. & van Wissen, L. (2011) An analysis of trends in spatial mobility of Dutch graduates,
Spatial Economic Analysis, 6, 5782.
Wheeler, C. H. (2006) Cities and the growth of wages among young workers: evidence from the NLSY, Journal
of Urban Economics, 60, 162184.
Yankow, J. J. (2006) Why do cities pay more? An empirical examination of some competing theories of the urban
wage premium, Journal of Urban Economics, 60, 139161.
419
Appendix. Data Restrictions

First of all, we remove graduates for which we lack pertinent information, such as work municipality, residential
municipality, university municipality or municipality at age 18 since we cannot discern complete spatial
movements for these individuals over time. Furthermore, we drop those individuals for whom we lack highschool grades or occupational codes since these are necessary for most analyses. To further balance the sample, we
remove those that die (very few) or who emigrate for some years of the period studied (874 individuals). A
mitigating factor in excluding these graduates is that they may be the ones with the highest observed and
unobserved ability, and excluding them could hence lead to a downward-bias on both the initial wage premium
and the wage-growth premium (cf. Venhorst et al., 2010). It is therefore unclear which effect dominates, and
hence in what direction a potential bias would run.
Table A1. Variable definitions

Variable
Wage
Age
Urban regions
City regions
Countryside regions
Social sciences
Natural sciences
Engineering
Average high-school grade
Works in MNE
Private sector
Establishment size
Tenure
Job change
Sector affiliation
Occupational codes
Classmates in residential
region with a job
Definition
The total wage earnings of a worker during a year. The exchange used is the
average exchange rate from 8 February 2012 to 8 February 2013
(1 EUR = 8.6792 SEK). The variable used in regressions is in logarithms
In full years, as of 1 January
A dummy for whether the graduate works in a region classified as urban
A dummy for whether the graduate works in a region classified as a city region
A dummy for whether the graduate works in a region classified as countryside
A dummy which is 1 if the graduate has a degree in the social sciences
A dummy which is 1 if the graduate has a degree in the natural sciences
A dummy which is 1 if the graduate has a degree in engineering
A dummy which is 1 if at least one of the parents have a long university
education (3 years)
A dummy which is 1 if the graduate has a spouse without a long university
education (3 years)
A dummy which is 1 if the graduate has a spouse with a long university
education (3 years)
A continuous variable between 1 and 5, where 1 is the lowest and 5 is the
highest. This serves as a proxy for ability
A dummy which takes the value 1 if the graduate has immigrated to Sweden
A dummy which takes the value 1 if the graduates parents have immigrated to
Sweden
A dummy which takes the value 1 if the graduate works in a MNE
A dummy which takes the value 1 if the graduate works in the private sector
The size of the establishment the graduate is employed at in November (in
logarithms)
The number of years the graduate has been employed at the current workplace.
Maximum tenure is the observational year minus 2001
A dummy which takes the value 1 if the worker changed occupation between
year t and t 1
Dummies for different sectors at the level of two-digit NACE sectors
Dummies for different occupations. From ILO-classifications on the one-digit
level
Parent with a manager or legislative position according to ILO-classifications
The number of former classmates (matched on university, level and field of
education as well as graduation year (either 1999 or 2000) who have a job
according to employment definitions and who reside in the same functional
region as the graduate
Copyright of Spatial Economic Analysis is the property of Routledge and its content may not
be copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for
individual use.

Market Thickness and The Early Labour Market Career of University Graduates

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Market Thickness and The Early Labour Market Career of University Graduates

Transféré par

Droits d'auteur :

Formats disponibles

Spatial Economic Analysis, 2014

Vol. 9, No. 4, 396419, http://dx.doi.org/10.1080/17421772.2014.961534

Market Thickness and the Early Labour Market Career

LINA AHLIN, MARTIN ANDERSSON & PER THULIN

paisseur du march initial de lemploi pour les diplms universitaires un

Market Thickness and the Early Labour Market Career

La densidad del mercado y el inicio de carrera en el mercado laboral de los

agglomeration economies; mobility; job switching

Market Thickness and the Early Labour Market Career

2. Thick Markets for Skills, City Size and Career Opportunities of

Market Thickness and the Early Labour Market Career

3. Data, Variables and Descriptives

3.2. Variables and Descriptive Statistics

Market Thickness and the Early Labour Market Career

4. Location, Wage Dynamics and Labour Market Mobility of University

Large urban regions

University educated workers

Table 3. Migration patterns of university graduates in 2000

The location choice and corresponding migration patterns of graduates may

4.2. Wages, Wage Growth and Job Switching

Market Thickness and the Early Labour Market Career

Mean yearly wage growth (%)

5. Identification Strategy and Results

The dependent variable, Ui, is a dummy variable which is 1 if university

Market Thickness and the Early Labour Market Career

location is simply not an artefact of that better graduates self-select to bigger

to these two main instruments, we also include university dummies as instruments.

Market Thickness and the Early Labour Market Career

Married, spouse with HE

Market Thickness and the Early Labour Market Career

Urban region (residential)

Establishment size (ln)

Urban region (work)

Market Thickness and the Early Labour Market Career

that switching to the private sector or to a multinational corporation raises wages

Urban region (work) all years

Active movers (outcome)

Market Thickness and the Early Labour Market Career

Market Thickness and the Early Labour Market Career

Market Thickness and the Early Labour Market Career

Appendix. Data Restrictions

Table A1. Variable definitions

Vous aimerez peut-être aussi